the
ECRM autokon 8400

The ECRM Autokon laser scanner/camera was a remarkable and ambititous device that used an extraordinarily complex hybrid analog computer, complex hard wired digital logic functions, a large optical bench and up to two helium-neon gas lasers to scan continuous tone monochrome photographs and directly write to bromide the half tone screened image. I learned to service these troublesome devices whilst working at Xenotrons' Melbourne HQ in 1983. This article is an attempt to lay some ghosts, this machine was one of the banes of my life at this time and the overwhelming support work load thus engendered, caused me to look for greener pastures. The machine contained many concepts that were new and unique and many examples of how not to do things. There seems to be absolutely no reference to this machine on the world wide web apart from mention in assorted operators resumes, even the parent company, ECRM, now fails to mention them much. They were a revolutionary machine for the time. They heralded great and profound changes occuring in the printing industry.

A traditional printing press is a "digital" device , it can either desposit a dot of ink, or not deposist a dot of ink. If it be monochrome, then that dot is black...and nothing else. A press cannot print even one tone of grey, let alone the thousands of grey levels that come with a monochrome continuous tone photograph. A means must be found to translate grey levels into discrete dots of ink, whereby the relative size of the  dot represents a grey level of the original image.  This is traditionally done in a large process camera, the original image is overlayed with a screen ruled at a resolution that is just printable by the printing process used by the publisher. So for example, cheap letterpress newsprint, the screen would be 55 lines per inch, and good offset lithographic printers, could use a screen of 100 lines per inch or better. The process camera takes , on high contrast high gain medium an image through this screen. If the gray level just visible through the grid is greater than the bromides' theshold, it is registered as a black picture element, if not it is considered white. The traditional process is expensive in materials, time and labor. The process camera operator was skilled in dark room techniques, it was a skilled trade and jealously protected by the craft unions. A good camera operator could do basic image enhancements like shift the contrast and brightness levels from a badly exposed original up to the press requirements. A truly skilled camera man could do unsharp masking to improve the subjective image resulution, but Im sure given the pernickety requirements, was only done by Intelligence organizations attempting to recover the maximum millitary value of impossible to obtain images. The pre press camera room would also require a library of screens for differant presses and processes, including novelty screens for quality magazines and books. These screens were hideously expensive , being created by special ruling engines. As I recall in 1983, these was supposedly only one producer of screens left on the planet and they charged like they meant it!

It was a large free standing machine, that had many internal improvements over the life of the product, but the basic shape and frame remained unchanged. There was a flat vacuum conveyor belt that held the original, moveable guides that performed the cropping function, a cryptic control panel of two parts, the normal operator controls and behind a door, the so called advanced controls. The conveyor brought the image into the scan station where two differant laser beams scanned the image, and as it scanned it wrote the "screened" image directly to bromide which would be discharged into a light proof canister. The original would be discharged at the end of the conveyor, if it had not snagged on the machine internals, thus requiring a field service call to remove it. The operator could set the input white and saturated black level on the machines front panel, the readin being derived from a densitometer, or by shear dumb luck. The midtone setting was done by guesswork, the effect of the control was to adjust the gamma of the net optical transfer function. In "lineart" mode the midtone setting changed the black/white decision threshold level. The operator selected the screen to be used and fiddled with controls to achieve artistic effects, provided that they understood the machine sufficiently. In my experience the operators never used the special features, they were not reproducible enough to form a work flow process around. In general they were just gratefull to get an image out of it before the machine broke! The machine was really intended to take the guesswork and black art of the camera room and render it into a "just push the button" semi skilled process, which naturally appealed to large newspaper publishers.  There was some saving in materials, as the Autokon machine generally had reproducible results, only one run was generally required to get a good screened image, compared with multiple runs on the process camera from which the "best" would be selected. The Autokon could also do anamorphic imaging, and this was heavily used by newspaper pre press, particularly be the advertiseing copy people that needed to make advertiseing artwork and type from external agencies who did not understand the newspapers format requirements, fit into the the newspapers column format. This was something that the traditional process camera could not do.

A large half inch thick aluminum plate held all the optics. Coherent light from two laser sources were directed and manipulated to create up to four seperate beams. One beam was directed to the scan station where it would scan the input image in one axis as it was being moved by the conveyor belt. Another beam was directed to an optical grating bar that would provide an internal beam position and reference clock . Another beam was created for writing to the recording film . One extra beam was created that was also directed to the scan station. This beam was defocused with respect to the other reading beam and also modulated at about 10Mhz. The combined video signal was processed to seperate the DC component the the modulated component. The two signals were processed at video speed in an analog computer to perform the level shifting functions and to perform unsharp masking at the same time. This part  was absolutely remarkable. The Autokon was famous for producing (when it worked!) crisp detailed halftones with "clarity" that sports editors loved.  In the original design two gas laser tubes and there was a troublesome servo using the acousto-optic modulator to modulate the intensity of one of the lasers. This caused so much trouble, especially as laser tubes did not age in the same way, that  design was abandoned and one laser with a beam splitter to derive the second laser source on the same optical bench was then used. When I took over Autokon maintenance in Australia I insisted that all Autokon had to be retrofitted with the single laser option. I was very glad for this change, it significantly reduced service calls. It seemed bizarre to me why two lasers had to be used given the extraordinary difficulty in getting two gas lasers to track each others' intensity. The scanning beam was deflected not by a spinning prism, as was done in laser typesetters and laser printers, but by a galvanometer motor, which was cheaper and required simpler optics.

The scanner station optical pickup was truly remarkable. Two offset ellipsoidal mirrors ran across the scan. The focal plane was lined with silicon photo voltaic diodes, the other focal plane was positioned so that the input image plane coincidended with it. The laser dot line scattering off the input image would be nicely focused onto the photodiode line. Very clever.  

The film output was driven by a belt driven synchronous motor that ran at a fixed (mains) fequency. Anamorphic reproduction was achieved in one axis by changing the speed of the vacuum conveyor belt.  A turret holding tiny mirrors deflected the four beams onto the galvanometer scanning mirror, all beams were directed through a field flattening lens that maintained tight focus over the scan plane.

The mirrors and beam splitters used silvered uncoated mirrors on aluminum mounting posts. This was the undoing of the machine. How long do you think uncoated silver stands up to the fumes in a camera room ? The Autokon was invariably placed inside damp, humid, dusty ,dark camera rooms and how long do you think silver mirrors lasted ?  Not very long. In a few months the silver coating would simply flake off the mirrors and along I would come to perform a complete optical alignment and mirror replacement. The aluminum posts did not help, they deflected, aged badly and adjustment threads would be stripped in no time. Later Autokons had steel posts. I dont know if coated mirrors were ever used, I did not stay around long enough to find out. In order to have the machine manufactureable, alignment posts and points were provided on the bench, and this was a very good part of the optical design, it could be completely aligned out in the field. I used to carry a small piece of polaroid film around with me, that way I could knock the laser intensity down to a level such that I could bear to look at the alignment targets. The laser tubes were nominally rated at 20milliwatts output and the machine was serviceable down to about 5 milliwatts. I spent a lot of time replacing gas lasers.

An acousto-optic modulator is a device for modulating gas lasers which cannot be modulated eletrically. It is a piezo electric device whereby a acoustic standing wave is created in the crystal (lithium niobate) by ahigh power radio frequency signal. This sets up regular density
waves in the cystal that has the same effect on coherent light as would a diffration grating. The AOM is in effect a programmable differation grating. The first order diffracted beam was the one that was used. This meant that 80 % of the laser power was thrown away, that was too bad, the need to modulate it was what was important. The efficiency of the process could be vagualey controlled by the power of the radio frequency energising source, and it was through the AOM that the heroic  attempt to make two gas lasers intensity track each each was acheived.  The method was never truly successfull, so ultimately, only one laser was used.  

The signal picked up by the scan station photo diodes was amplifield in a circuit that went through many iterations. The amplified siganl went to the analog board. The sharp beam signal was seperated from the unsharp beam signal. The unsharp signal went through bandpass amplifiers and a synchronous detector to recover the actual video signal.  A amplitude differance signal was produced between the sharp and unsharp video.  A variable amount of the differance signal could be added to the final output video signal. This had the effect of enhancing the edges of images and improved edge contrast. Programmable amplifiers set the gain and offset of the video. A sub feed of the video went to a network of analog multipliers and high speed op amps that performed the midtone manipulation and the effect was to perform a variable gamma operation whereby the mid point of the optical transfer function could be shifted in shape, within the bounds set by the highlight/shadows endpoint controls. This analog computer had a long and torterous adjustment procedure with many interacting pots. You never really knew, at the end, if you had done it right. The most galling aspect to this procdure was that the analog board was supported only on its edge connectors which carried millivolt level analog signals. The adjustment was sensitive to flexture in the edge connectors. This was absolutely revolting as the ajustment had to be carrier out on an extender board. The analog board also contained the bit slicer, a high speed comparator that compared the instantaneous video level level with the output of the "screen" rom to determine if a pixel element was to be set on the film recorder.

The dot board was what actaully performed the "screening" function. The screen roms contained an image of an actual screen. They were addressed by a counter whose clock source was derived from a phase locked loop whose reference clock was the signal coming from a phtodiode strip behind an optical grating that was scanned by one of the four beams. Another phase locked loop was driven by the conveyor belt motors optical shaft encoder. This also fed another counter. Here we had two counters, whose count was a function of instantaneous beam position and conveyor belt position, these addressed the screen rom whoose outputs fed a high speed DAC. The output of the DAC could be shifted in gain and offset ( the bane of Autokon engineers) and the output compared with that comparator in the video board. Here was the mechanism that converted the continous tone input video into the half tone signal which then on/off modulated the recording laser across the bromide as it scanned and moved in the vertical axis. The dot board was a digital nightmare, there was no clear design. It had clearly evolved from kludge upon kludge and I recall being told that nobody really understood it, the design worked and while it worked they would continue to manufacture it! These days we would use an FPGA or high speed DSP or even a network of PALs. This was all asynchronous random logic WITH feedback. Foretuneately none of these horrible circuits died under my watch, just as well.

The entire machine was controlled by a network of 555 timers, most likely  designed by MIT vacation student engineers. In an era when microprocessors were around ( think 6800, 8080 et al ) boards covered with trimpots, 555s and random logic sequenced all functions of this machine. Setup was an immensely frustrating affairs of trial and error as we wrestled with inumerable trimpots to make the machine behave correctly. The controller boards were not interchangeable without going through the tuneup procedure. One of the most objectionable features were the cropping arms. The design and intention was good. The cropping cursors were attached to 10 turn servo potentionmenters, the output gained and offsetted to internal meaningfulness. An ADC converted the x axis to digital values that were compared with the x axis counters. This would define the "cropping" by enabling the print beam inside the crop area. The y axis crop was set by comparators that determined when the conveyor leading edge reached the scan station and when it left. This settup could take a day or trial and error, most times it was about right. After all, cropping could be done with a Stanley  Knife ("box cutter") far more effectively !

Then there was the strange autocal board.  One of the problems with the scan station photodiode sensor strip was that it used photo voltaic silicon photo diodes. These were not very sensitive at the red He-Ne laser light frequency, and their sensitivity is not constant between diode modules and even not constant over the active area of the sensor.  This would require carefully matching of photo-diode units in production. At worst it could lead to a vertical banding in the output image. That was not all. The camera operators would take output from the compositors that used hot wax to cut and paste there images down to board. Over time, vapours from the wax would condense on the cold elliptical mirror surfaces and this would result in banding on the output image. The autocal board would be primed with an alternating carboard pattern of saturated white and saturated black, the density being read of a densitometer, the black and white endpoint controls set. The machine would read in the the chart and then fill a RAM with gain and offset control data that would be dynamically applied to the recovered video signal. When the autocal board worked it could automatically compensate for most of the machines self induced drift. As I recall, autocal failures were incredibly diffcult to debug.

There was an optional screens board, containing a bank of ROMS containing optional screens, including standard ruled screens and a large variety of artistic screens, inlcuding the fascinating greek maze and the remarkable mezotint screen. The art of the mezotint, was commented on by the great M.C. Escher as being impossibly difficult to accomplish by traditional methods, yet the Autokon (when it worked!) could do a mezotint like the masters.

The online market had not been overlooked, there was the much cursed Formatter Board. It was a unique and irreproducible hi speed buffered parallel interface much cursed by interface designer engineers at Xenotron and Monotype .  I saw an interface board designed by Linotype for their system, it plugged directly into the Autokons backplane and utilized the raw timing and analog signals. The Autokon could be used as an online scanner, and could deliver to a host system raw contone data, or screened halftone data.  ECRM had recognised the online market and also marketed a cutdown version of the Autokon without some of the output recording components as the model 8200 (later the 720)

I had to support a small number of these monsters. They were effectively unsupportable in the field. It invariably become a finger pointing squabble with the host computers vendor as to which one was busted. Without any kind of stand alone visualization tool, all we could do was examine various test points for known waveforms and see that the online signals going to the host were wiggling up and down in a sensible way. Funnily enough, the faults that I encountered were mostly the fault of assorted stand alone interface boxes, full of random logic and microprocessors that attempted to take the prosiac formatter board ouput and turn it into something that third party systems could understand, like GPIB. ( GPIB was to controllers and minicomputer what USB is today ). I seem to recall that the formatter board interface was bi-directional, the host system could write and image to the Autokon for output, but I never saw anybody use this feature, which was a pity because it would have been a very capable output engine for a RIP.

Eventually I solved the problem of hum getting into the low level video. It was nothing more than re routing the main video amplifiers power cabling so that it followed the same route as a coaxial cable carrying the video signal, thus eliminating a hum loop being threaded by the stray field from the power supplies. I performed this fix before ECRM's engineering manager at the time, I dont know if the fix ever went into production, not-invented-here syndrome probably lost this cause.